I'd go the complete opposite. If the data requiresa unique index (for data quality), then go ahead and create the index. I wouldn't do a unique index just because it might go faster and not consider wheter or not it could bug the applications using it.